Pr

1
Using Altimetry Waveform Data and Ancillary
Information from SRTM and LANDSAT to retrieve
River Characteristics By Vivien ENJOLRAS1) and
Ernesto RODRIGUEZ(2) (1) CNES/LEGOS, 18, av.
Edouard Belin, 31400 Toulouse, France (2) Jet
Propulsion Laboratory, 4800 Oak Grove Drive,
Pasadena CA 91109, United States Author to whom
correspondence should be addressed - Email
vivien.enjolras_at_gmail.com
Main Philosophy
Preliminary Scene File Creation
Abstract
Spaceborne radar altimeters are shown to have the
potential for monitoring the height of large
rivers with accuracies of approximately 1 m.
However, the need for a better height accuracy
and the observations of smaller continental
basins have led to studies on how to improve and
extend the use of nadir altimeter data.
Conventional retracking techniques over land are
limited to the examination of altimeter waveforms
on a case by case basis. Due to the arbitrary
geometry which may be present at altimeter river
crossings, this approach may be limited to large
rivers, which approximate ocean crossings. To
overcome this limitation, we introduce a waveform
fitting method which uses the entire set of
waveforms associated with a water crossing,
rather than individual waveforms. By using
ancillary data, such as a digital elevation model
(obtained from SRTM, gtopo30) and classification
maps (obtained from Landsat, Modis), it is
possible to recast the retracking problems as a
maximum likelihood estimation problem.
Theoretical power returns based on the a priori
knowledge of the observed scenes are generated
resulting in a parametric library of waveform
histories, which is then used to constrain the
estimation. For demonstration, we concentrate on
the river Meuse in Northern Western Europe, and
on the river Lena in Russia. The Meuse has
important social impact, since it has flooded in
the past and better real time predictions of its
changing stage may improve flood forecasting
skill. Furthermore, it presents a challenge to
conventional nadir altimeter waveform retracking.
We will present both theoretical performance
results and demonstrate the feasibility based on
real altimeter data. Keywords altimetry,
waveform, hydrology, Europe

• Current Methods can be qualified of Single
  Waveform Processing
• Waveforms classification (11 classes on ENVISAT)
  with specific algorithms of estimation
• Method of estimation by deconvolution
• Neuronal network
• We on the contrary consider using the whole
  waveform history over continental water areas
  (more information)
• Use of ancillary data (information added)
• Optic for scene masks LANDSAT, MODIS
• Topography SRTM, Gtopo30

Estimation Process Theory
Developed Program Chart

• Simulated Power Returns are function of time,
  instrument parameters (wavelength, power, antenna
  pattern, point target response ?), the satellite
  motion (range, incidence), and the observed scene
  (backscattering coefficient), as followed
• Pulse compression and onboard digital sampling
  of the returned signals at the frequency 1/T are
  equivalent to measure the time history of the
  return in a sequence of range gates separated by
  an effective two-way travel time resolution of
  1/B, or a range resolution of c/(2B) (47 cm for
  Topex)
• A template corresponds to a specific set of
  values for the parameters willing to be
  estimated
• Elevation of inland water above its a priori
  reference resulting from ancillary data
• Backscattering coefficient, standard deviation of
  small gravity-capillary waves
• River slope, etc
• Numerous templates with stepped sets of related
  scene parameters values can be generated
• A MAP search is performed to find the set of
  parameters values resulting in the closest
  template to the real data
• Assuming that the performed search has found a
  best template, close enough to the observed
  waveform history, its difference with the real
  data can finally be considered as linear in the
  parameters to be estimated. Considering the
  elevation and the backscattering coefficient to
  be estimated
• The derivatives can be computed for the best
  template set of values, and enables then to
  estimate the parameters with a maximum likelihood
  process

Theoretical Validation of the Method and the
Machine
Example of Results on rivers Lena and Meuse

• Three theoretical scenes, for different
  purposes, have been considered
• Ocean-like scene (validation of the power returns
  simulation)
• Very Bright Scene Crossing (validation of the
  scene generation and the power returns
  simulation)
• Two rivers whose separate distance, own height
  and own backscattering coefficient can be
  modified (Study of the parameters correlation)
• In the third study, the derivatives are
  calculated, and the covariance matrix of the
  parameters then informs on their correlations

• On narrow continental water areas, water
  templates generation is fast, but land is slower.
  The short ground track motion between each cycle
  (maximum 1 km on T/P) prevents from running land
  returns each time, as illustrated by the
  following correlation matrix of land returns for
  T/P cycles 380, 381, 384, 385 and 386 over Meuse

Conclusions and Perspectives
1,25 m above ancillary data basis Mean 20 days in
situ data around same day in 2005 1.2 m above
basis

• Theoretical Results support the new hopes
  offered by such a method parameters correlation
  are low which shows that this 2D approach enables
  to distinguish narrow and close inland waters
• First Results over Meuse and Lena are very
  convincing and favour the continuity of such a
  study at higher processing scales
• Topex Waveforms are very peaky on Meuse and
  Lena, showing a very specular behaviour. The
  model of backscattering coefficient may need to
  be improved and detailed to get closer to real
  data in simulations
• One year data is about to be processed over the
  Lena using MODIS data (masks computed by Dr.
  Larry Smith Team in University of California, Los
  Angeles) to get accurate evolving water masks

Real Altimetry Data Processed

• Topex data have been firstly selected as the
  tracker behavior on Topex is much smoother than
  on Poseidon2, resulting in a lot more data on
  inland waters
• Topex SDR (Sensor Data Record)
• 10 Hz Ku-band waveforms (64 range gates)
• 20 Hz Range Ku tracker behavior
• 20 Hz AGC automatic gain control behavior (power
  onboard attenuation)
• 1 Hz latitude and longitude
• Topex GDR (Geophysical Data Record)
• 10 Hz satellite altitude above reference
  ellipsoid
• 1 Hz flag
• 1 Hz dry troposphere correction

References
Berry P. A. M. Garlick, J. D. Freeman, J. A.
Mathers, E. L. Global inland water monitoring
from multi-mission altimetry. Geophysical
Research Letters 2005, 32. Birkett, C.M. Mertes,
L.A.K Dunne, T. Costa, M. Jasinski, J.
Altimetric remote sending on the Amazon
Application of satellite radar altimetry. J.
Geophysic. Res. 2002, 107 (D20), 8059,
10.1029. Chelton, D. B. Walsh, E.J. MacArthur,
J. L. Pulse compression and sea level tracking in
satellite altimetry. J. Atmos. Oceanic Technol.
1989, 6, 407-438. Frappart, F. Calmant, S.
Cauhope, M. Seyler, F. Cazenave, A. Preliminary
results of ENVISAT RA-2-derived water levels
validation over the Amazon basin. Remote Sensing
of Environment 2006, 100, 252-264. Rodriguez, E.
Chapman, B. Extracting ocean surface information
from altimeter returns the deconvolution method.
J. Geophysic. Res. 1989, 94 (C7), 9, 761-9, 778.
15 Years of Altimetry, Venice, March 2006